A pressure balance-based driving face ventilation equipment linkage control method and device

Abstract

The application provides a pressure balance-based linkage control method and device for a driving face ventilation equipment. The method of the application aims to provide an intelligent dust collection system, which obtains environmental parameters of a driving face in real time through various sensors such as dust, temperature and wind speed, and realizes linkage control of the wind speed, air volume and wind direction of a wind distributor, a dust hood and a negative pressure air duct in combination with data analysis and an automatic control system. The method can effectively improve the dust capture efficiency, enhance the pressure balance capacity of the system, significantly reduce the dust concentration of the working face and optimize the temperature environment, and is also helpful to reduce energy consumption and ensure the safety of operation.

Description

Technical Field

[0001] This invention relates to the field of control technology, and in particular to a method and device for linkage control of ventilation equipment in tunnel faces based on pressure balance. Background Technology

[0002] Ventilation and dust control systems at coal mine tunneling faces are a core technology for ensuring underground operational safety and are widely used in the coal mining industry. With the increasing demand for intelligent mine construction, existing technologies have evolved from traditional manual experience-based control to automated monitoring, but the limitations of isolated equipment operation remain. In related technologies, a basic environmental parameter acquisition network is constructed through the collaborative operation of dust sensors, temperature sensors, and wind speed sensors, combined with PID control algorithms to achieve single-point parameter adjustment. Specifically, this system covers the entire process from data acquisition to control execution, including key aspects such as airflow direction guidance, air volume distribution optimization, and pressure difference compensation. The independent control mode of the air distributor, dust collection hood, and negative pressure ventilation duct has become the mainstream architecture in the industry.

[0003] However, existing ventilation and dust control methods rely directly on single sensor data and human experience for decision-making, without establishing a dynamic coordination mechanism between multiple devices. This can lead to systemic defects such as localized airflow imbalances and low dust collection efficiency. Specifically, there is a disconnect between the parameters of the air distributor angle adjustment and the dust collection efficiency of the dust hood; the airflow adjustment of the negative pressure duct does not consider the wind resistance disturbance caused by changes in the air distributor opening angle; and traditional PID control only provides feedback adjustment for a single parameter, failing to handle the complex dynamic interactions between multiple devices. Consequently, human experience-based adjustments exhibit significant response delays in scenarios with sudden changes in operational intensity, such as blasting and coal cutting. Furthermore, existing algorithms have inherent limitations in terms of control accuracy and energy consumption balance, resulting in equipment operating at full load for extended periods. Among these, the delayed early warning of excessive dust concentration and the risk of gas accumulation caused by abnormal temperatures have become key technical bottlenecks restricting the improvement of coal mine safety and production efficiency. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to propose a linkage control method for ventilation equipment in tunneling faces based on pressure balance.

[0006] The second objective of this invention is to propose a linkage control device for ventilation equipment in tunnel faces based on pressure balance.

[0007] To achieve the above objectives, the first aspect of this invention proposes a method for coordinated control of ventilation equipment at a tunnel face based on pressure balance, comprising: S1, synchronously collecting multi-source environmental parameters and equipment operating parameters related to the tunnel face; S2, after standardizing the collected parameters to a unified range, calculating dust concentration deviation, temperature deviation, and pressure difference deviation based on a preset safety threshold, and constructing a feature dataset including dust change rate, airflow pressure difference, and work intensity coefficient; S3, using the feature dataset and the calculated deviation data to generate a basic adjustment amount through a PID control algorithm, and performing dynamic weight correction by combining the optimal parameter combination output by the machine learning model, to obtain coordinated control parameters for the airflow direction adjustment amount of the air distributor, the dust collection efficiency optimization amount of the dust hood, and the airflow adjustment amount of the negative pressure duct; S4, using an LSTM time-series prediction model to analyze historical and real-time data, predicting the dust concentration and temperature change trends, and when the predicted value exceeds the safety threshold, calculating the advance control amplitude based on the prediction deviation and generating a pre-control instruction based on the coordinated control parameters as the execution value, synchronously adjusting the airflow of the negative pressure duct and the opening angle of the air distributor to achieve airflow balance.

[0008] In one embodiment of the present invention, the synchronous acquisition of multi-source environmental parameters and equipment operating parameters related to the tunnel face further includes: S11, using dust sensors, temperature sensors, wind speed sensors and pressure difference sensors for distributed point acquisition, with an acquisition frequency of once every 1 minute; S12, uploading the acquired parameters to the data analysis system in real time through a wireless transmission module, and using edge computing nodes for preliminary data fusion processing.

[0009] In one embodiment of the present invention, after standardizing the collected parameters to a unified range, calculating the dust concentration deviation, temperature deviation, and pressure difference deviation based on a preset safety threshold, and constructing a feature dataset including dust change rate, airflow pressure difference, and work intensity coefficient, further includes: S21, mapping parameters such as dust concentration, temperature, and wind speed to the [0,1] range using a linear normalization method to eliminate dimensional differences; S22, selecting dust change rate, airflow pressure difference, and work intensity coefficient as key features, and performing feature dimensionality reduction processing using principal component analysis.

[0010] In one embodiment of the present invention, the step of generating a basic adjustment quantity using a PID control algorithm based on a feature dataset and calculated deviation data, and then dynamically correcting the weights by combining the optimal parameter combination output by the machine learning model to obtain the coordinated control parameters for the airflow direction adjustment of the distributor, the dust collection efficiency optimization of the dust hood, and the airflow adjustment of the negative pressure duct further includes: S31, dynamically adjusting the weight ratio of the PID control quantity and the output parameters of the machine learning model according to the current work intensity coefficient; the higher the work intensity, the greater the weight ratio of the machine learning model; S32, performing nonlinear compensation on the corrected coordinated control parameters using a fuzzy logic algorithm to eliminate control errors caused by equipment response delay.

[0011] In one embodiment of the present invention, the step of using an LSTM time-series prediction model to analyze historical and real-time data, predict dust concentration and temperature change trends, and when the predicted value exceeds a safety threshold, calculating the advance control amplitude based on the prediction deviation and generating a pre-control instruction based on the collaborative control parameters to simultaneously adjust the negative pressure duct air volume and the air distributor opening angle to achieve airflow balance further includes: S41, constructing a neural network model containing two layers of LSTM units, wherein the first layer contains 64 neurons and the second layer contains 32 neurons, and setting a 0.2 ratio Dropout layer between hidden layers to prevent overfitting; S42, using a 3-time-step moving average algorithm to denoise the preprocessed time-series data, while using linear interpolation to complete short-term missing data and using the average of the previous 30 minutes to replace long-term missing data.

[0012] In one embodiment of the present invention, the method further includes: S5, when the system detects that the global pressure balance coefficient is in a state of severe imbalance, a three-step coordinated control process is executed: S51, the local negative pressure is increased by increasing the air volume of the dust collection hood first; S52, the frequency of the negative pressure duct motor is dynamically adjusted according to the total air volume requirement; S53, the opening angle of the air distributor is finely adjusted to calibrate the remaining air pressure deviation, and the above three-step process is repeated within the feedback cycle until the system pressure is balanced.

[0013] To achieve the above objectives, a second aspect of the present invention proposes a pressure balance-based linkage control device for ventilation equipment in a tunnel face, comprising: a multi-source parameter acquisition module for synchronously acquiring multi-source environmental parameters and equipment operating parameters related to the tunnel face; a parameter standardization and feature construction module for standardizing the acquired parameters to a unified range, calculating dust concentration deviation, temperature deviation, and pressure difference deviation based on a preset safety threshold, and constructing a feature dataset including dust change rate, airflow pressure difference, and work intensity coefficient; and a PID and machine learning collaborative control module for using the feature dataset and calculated deviation data to implement PID control. The control algorithm generates basic adjustment values, and combines the optimal parameter combination output by the machine learning model for dynamic weight correction to obtain the coordinated control parameters for the airflow direction adjustment of the distributor, the dust collection efficiency optimization of the dust hood, and the airflow adjustment of the negative pressure duct. The LSTM time series prediction and pre-control module is used to analyze historical and real-time data using the LSTM time series prediction model to predict the trend of dust concentration and temperature changes. When the predicted value exceeds the safety threshold, the pre-control amplitude is calculated based on the prediction deviation, and a pre-control instruction based on the coordinated control parameters is generated to simultaneously adjust the airflow of the negative pressure duct and the opening angle of the distributor to achieve airflow balance.

[0014] In one embodiment of the present invention, the multi-source parameter acquisition module is further configured to: use dust sensors, temperature sensors, wind speed sensors, and pressure difference sensors for distributed point acquisition, and acquire relevant multi-source environmental parameters and equipment operating parameters, wherein the multi-source environmental parameters include: dust concentration, temperature, wind speed, pressure difference, and the opening angle of the air distributor, the dust collection efficiency of the dust collection hood, and the air volume of the negative pressure duct; and upload the acquired parameters to the data analysis system in real time through a wireless transmission module, and perform preliminary data fusion processing using edge computing nodes.

[0015] In one embodiment of the present invention, the parameter standardization and feature construction module is further used to: map dust concentration, temperature and wind speed parameters to the [0,1] interval through linear normalization to eliminate dimensional differences; screen dust change rate, airflow pressure difference and work intensity coefficient as key features, and perform feature dimensionality reduction processing using principal component analysis.

[0016] In one embodiment of the present invention, the PID and machine learning collaborative control module is further configured to: dynamically adjust the weight ratio of the PID control quantity and the machine learning model output parameters according to the current work intensity coefficient, wherein the higher the work intensity, the greater the weight ratio of the machine learning model; and perform nonlinear compensation on the corrected collaborative control parameters through a fuzzy logic algorithm to eliminate control errors caused by equipment response delay.

[0017] In one embodiment of the present invention, the LSTM time series prediction and pre-regulation module is further configured to: construct a neural network model containing two layers of LSTM units, wherein the first layer contains 64 neurons and the second layer contains 32 neurons, and a Dropout layer with a ratio of 0.2 is set between the hidden layers to prevent overfitting; use a 3-time-step moving average algorithm to denoise the preprocessed time series data, and use linear interpolation to complete short-term missing data and mean replacement to replace long-term missing data.

[0018] In one embodiment of the present invention, it further includes: a pressure imbalance collaborative control module, used to execute a three-step collaborative control process when the system detects that the global pressure balance coefficient is in a severely unbalanced state: firstly, increase the local negative pressure by increasing the air volume of the dust collection hood; dynamically adjust the frequency of the negative pressure duct motor according to the total air volume requirement; fine-tune the opening angle of the air distributor to calibrate the remaining air pressure deviation, and repeat the above three-step process within the feedback cycle until the system pressure is balanced.

[0019] The method and apparatus of this invention can realize the coordinated control of the air distributor, dust collection hood and negative pressure duct at the tunnel face, effectively improve dust collection efficiency and system pressure balance capability, reduce energy consumption and enhance the prediction and active control level of dust concentration changes.

[0020] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0022] Figure 1 This is a flowchart of a method for linkage control of ventilation equipment in a tunnel face based on pressure balance, according to an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of a pressure balance-based ventilation equipment linkage control system for a tunnel face according to an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of a pressure balance-based ventilation equipment linkage control device for a tunnel face according to an embodiment of the present invention. Detailed Implementation

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0027] The following description, with reference to the accompanying drawings, describes a method and apparatus for linkage control of ventilation equipment in a tunnel face based on pressure balance, according to an embodiment of the present invention.

[0028] Example 1

[0029] Figure 1 This is a flowchart of a method for coordinated control of ventilation equipment at a tunnel face based on pressure balance, according to an embodiment of the present invention. Figure 1 As shown, it includes:

[0030] S1 synchronously collects multi-source environmental parameters and equipment operating parameters related to the tunneling face.

[0031] Specifically, this step of "synchronously collecting multi-source environmental parameters and equipment operating parameters such as dust concentration, temperature, wind speed, pressure difference, opening angle of the air distributor, dust collection efficiency of the dust hood, and air volume of the negative pressure ventilation duct" is the basic link to realize intelligent ventilation and dust control at the tunnel face. Its technical implementation relies on a multi-sensor fusion acquisition and data synchronization mechanism.

[0032] Specifically, the system adopts a distributed sensor network architecture, and the sensors deployed at key locations on the tunnel face include laser scattering dust concentration sensors (range...). The system includes sensors with accuracy ±10%), infrared temperature sensors (range -20℃~80℃, accuracy ±0.5℃), hot-wire anemometers (range 0-30 m / s, accuracy ±2%), and differential pressure sensors (range 0-1000 Pa, accuracy ±1%FS). Meanwhile, the distributor angle is fed back in real-time by an encoder (resolution 0.1°), the dust collection efficiency of the dust hood is calculated using the rate of change in airflow and dust concentration, and the airflow of the negative pressure duct is measured by a turbine flow meter or ultrasonic anemometer (accuracy ±1.5%).

[0033] Furthermore, the parameter acquisition frequency is uniformly set to once per minute, and synchronous transmission of multi-source data is achieved through RS485 or CAN bus protocols to ensure the consistency of each parameter in the time dimension. The acquired data undergoes preliminary preprocessing at the edge computing node, including linear interpolation to complete short-term missing data, moving average denoising (such as 3-time-step moving average), and standardization processing (such as mapping wind pressure data to the 0-1 interval) to eliminate dimensional differences and improve the efficiency of subsequent algorithm processing.

[0034] Specifically, this step plays a crucial supporting role in the entire system, providing high-quality, real-time input data for PID control, machine learning model training, and LSTM prediction models. Through the synchronous acquisition of multi-source parameters, the system can comprehensively perceive the dynamic environmental changes at the tunnel face, providing a reliable data foundation for subsequent linkage control and predictive regulation, thereby achieving precise control of dust concentration and efficient coordinated operation of the ventilation system.

[0035] Furthermore, S1 includes:

[0036] S11 uses a distributed collection point system with dust sensors, temperature sensors, wind speed sensors, and differential pressure sensors, with a collection frequency of once every minute.

[0037] Specifically, this step involves distributed data collection using dust sensors, temperature sensors, wind speed sensors, and differential pressure sensors, with a collection frequency of once every minute. This technology achieves a collaborative deployment and synchronous data acquisition mechanism based on a multi-source heterogeneous sensor network, aiming to provide highly timely and accurate environmental parameter inputs for the intelligent ventilation and dust control system at the tunnel face, thereby supporting the efficient operation of subsequent PID control, machine learning models, and LSTM prediction algorithms.

[0038] Specifically, sensor deployment follows the principle of "zonal coverage and priority for key nodes." Based on the geometry and airflow distribution characteristics of the tunneling face, sensors are deployed in key locations such as the intake side, return air side, coal cutting area, blasting area, and areas with high personnel activity. The dust sensor uses the laser scattering principle and has a measurement range of [missing information]. The resolution is The sampling period is 1 minute; the temperature sensor is a PT100 type, with a measurement range of -20℃ to +60℃ and an accuracy of ±0.5℃; the wind speed sensor is ultrasonic, with a range of 0.1-30 m / s and an accuracy of ±0.2 m / s; the differential pressure sensor is a differential pressure transmitter, with a range of 0-2000 Pa and an accuracy of ±1%FS. All sensors support Modbus or CAN bus communication protocols for easy integration into the system's data acquisition module.

[0039] Optionally, the system employs edge computing nodes for data preprocessing, including linear interpolation to complete short-term missing data, moving average denoising (such as 3-time-step moving average), and standardization (such as Min-Max normalization to the [0,1] interval) to improve data quality and model input stability. The acquisition frequency is set to once every 1 minute, which meets the real-time requirements of coal mine safety monitoring systems for key parameters such as dust concentration (such as the recommendations for dust concentration monitoring frequency in AQ 1029-2019 "Management Specifications for the Use of Coal Mine Safety Monitoring Systems and Detection Instruments").

[0040] Furthermore, this step provides fundamental support for the data closed loop of the entire intelligent control system. Through high-frequency, multi-dimensional environmental parameter acquisition, the system can accurately identify dust diffusion paths, abnormal temperature areas, and airflow imbalance points, providing deviation input for the PID control algorithm, building training datasets for machine learning models, and providing time-series input windows for LSTM prediction models. This enables dynamic prediction and early control of dust concentration, improves system response speed and control accuracy, and ensures the health and safety of operators.

[0041] S12 uploads the collected parameters to the data analysis system in real time via a wireless transmission module, and performs preliminary data fusion processing using edge computing nodes.

[0042] Specifically, this step involves uploading collected environmental parameters from the tunnel face (such as dust concentration, temperature, wind speed, and airflow pressure difference) to the data analysis system in real time via a wireless transmission module, and performing preliminary data fusion processing at the edge computing node. This process is a key step in realizing intelligent ventilation and dust control at the tunnel face, and its technical implementation is based on the Industrial Internet of Things (IIoT) architecture and edge computing technology to ensure low-latency data processing and efficient transmission.

[0043] Specifically, the wireless transmission module uses LoRaWAN or ZigBee protocols for data communication, featuring low power consumption, long-distance transmission, and strong anti-interference capabilities, making it suitable for the complex electromagnetic environment of underground coal mines. Sensor nodes sample at a 1-minute interval, encrypting the collected raw data and packaging it into data frames conforming to the MQTT protocol, which are then uploaded to the cloud or local edge server via a wireless gateway. During data transmission, CRC checksums and retransmission mechanisms are employed to ensure data integrity and reliability, with transmission latency controlled within 500ms to meet real-time control requirements.

[0044] Furthermore, edge computing nodes are deployed in edge servers or gateway devices near the tunnel face, running lightweight data fusion algorithms. These nodes receive heterogeneous data streams from multiple sensors, first performing data alignment and timestamp synchronization to ensure consistency of parameters over time. Subsequently, a weighted average method or Kalman filtering algorithm is used to fuse key parameters such as dust concentration and wind speed, eliminating sensor noise and measurement biases and improving data accuracy. For example, the dust concentration fusion error can be controlled within ±5%, and the wind speed measurement error does not exceed ±0.2 m / s.

[0045] Specifically, this step plays a crucial role in the system, providing high-quality, low-latency data input for subsequent PID control, machine learning model inference, and LSTM prediction models. Through preliminary processing via edge computing, the system can effectively reduce the computing burden on the cloud, improve response speed, and achieve dynamic prediction of dust concentration and coordinated control of equipment, thereby enhancing the overall system's intelligence and operational efficiency.

[0046] S2 standardizes the collected parameters to a uniform range, calculates the dust concentration deviation, temperature deviation, and pressure difference deviation based on preset safety thresholds, and constructs a feature dataset including dust change rate, airflow pressure difference, and work intensity coefficient.

[0047] Specifically, in some implementations, after standardizing the collected parameters to a unified range, dust concentration deviation, temperature deviation, and pressure difference deviation are calculated based on preset safety thresholds. A feature dataset containing dust change rate, airflow pressure difference, and work intensity coefficient is then constructed. This is a key step in this intelligent control method to achieve multi-source information fusion and coordinated equipment control. This step eliminates the dimensional differences between different sensor data through standardization, providing a unified and comparable data foundation for subsequent PID control, machine learning model training, and LSTM prediction models.

[0048] Specifically, firstly, the dust concentration (unit: The raw data, including temperature (°C) and airflow pressure difference (Pa), are standardized. The Min-Max standardization method is typically used to map the data to the [0,1] interval. After standardization, the system compares each parameter with preset safety thresholds to calculate the dust concentration deviation (ΔC), temperature deviation (ΔT), and pressure difference deviation (ΔP). The deviation value can be expressed as the absolute value or relative percentage of the difference between the current value and the threshold, depending on the control strategy settings.

[0049] Furthermore, the system extracts key features to construct a feature dataset. The dust change rate (ΔC / Δt) calculates the concentration change trend per unit time using a sliding window, reflecting the dynamic characteristics of dust diffusion or accumulation; the airflow pressure difference (ΔP) is used to evaluate the stability of the ventilation system and the efficiency of airflow distribution; the work intensity coefficient (WIF) quantifies the impact of the current operation on dust generation by analyzing the operating status of the tunneling equipment (such as coal cutting frequency, blasting cycle, etc.). When constructing the dataset, these features must meet requirements such as consistent sampling frequency (e.g., 1 minute / time), time alignment, and data integrity to ensure the accuracy and generalization ability of the model training.

[0050] Specifically, this step is widely used in intelligent ventilation systems at coal mine tunneling faces, especially in scenarios involving multi-device collaborative control, dynamic prediction, and adaptive adjustment. By constructing a high-quality feature dataset, the system can more accurately identify dust diffusion patterns and changes in ventilation demand, providing reliable data support for subsequent PID control, machine learning model inference, and LSTM prediction.

[0051] Furthermore, through standardization and feature extraction, the fusion efficiency of multi-source data and the interpretability of model input are significantly improved, laying a solid foundation for real-time control and predictive management of environmental parameters at the tunnel face. Simultaneously, this method effectively supports the system's adaptive response capability under complex working conditions, enhancing the intelligence level and operational efficiency of dust control.

[0052] Furthermore, S2 includes:

[0053] S21 maps parameters such as dust concentration, temperature, and wind speed to the [0,1] interval using a linear normalization method to eliminate dimensional differences.

[0054] Specifically, in this patented method, mapping parameters such as dust concentration, temperature, and wind speed to the [0,1] interval using a linear normalization method is a key step in achieving multi-source heterogeneous environmental data fusion processing. This step aims to eliminate the dimensional differences between the parameters, providing a unified input space for subsequent PID control, machine learning model training, and LSTM prediction model, thereby improving the convergence speed and control accuracy of the algorithm.

[0055] Optionally, to enhance the robustness of the system, a dynamic threshold update mechanism can be introduced during the normalization process. This involves dynamically calculating x_min and x_max based on a sliding window of real-time acquired data (e.g., the past 30 minutes) to adapt to parameter fluctuations caused by changes in the intensity of operations at the tunnel face. Furthermore, the normalized data can be used as the input deviation of the PID controller, the feature vector of the machine learning model, and the time-series input of the LSTM prediction model, ensuring that all algorithm modules work collaboratively on a unified scale.

[0056] Specifically, this step plays a fundamental role in the entire intelligent control method. By eliminating dimensional differences, the system can more accurately assess the influence weight of each environmental parameter on ventilation and dust removal effects, improving the scientific nature and real-time response capability of the control strategy. Simultaneously, normalization helps reduce the complexity of model training, improves prediction accuracy, and provides data support for achieving efficient, energy-saving, and intelligent ventilation control at the tunnel face.

[0057] S22, dust change rate, airflow pressure difference, and work intensity coefficient were selected as key features, and principal component analysis was used to perform feature dimensionality reduction.

[0058] Specifically, in this method, the system selects dust change rate, airflow pressure difference, and work intensity coefficient as key features, and uses principal component analysis (PCA) for feature dimensionality reduction to improve the computational efficiency and model generalization ability of subsequent intelligent control algorithms. This step plays a crucial role in the multi-source environmental information perception and intelligent control system, and is a key link in achieving efficient and low-energy dust control.

[0059] Specifically, in some implementations, the dust change rate is calculated using time series differencing, that is, the difference between the dust concentration at the current moment and the concentration at the previous moment divided by the time step (e.g., 1 minute), with units of... The dynamic trend of dust concentration is characterized by the air pressure difference, which is collected by the air pressure sensors before and after the air distributor. The difference is calculated and standardized to the [0,1] interval to eliminate the influence of dimensions and facilitate subsequent algorithm processing. The operation intensity coefficient is calculated by weighting the tunneling operation plan data (such as coal cutting, blasting, etc.) and the actual equipment operating parameters (such as tunneling machine speed and propulsion speed), reflecting the potential impact of the current operation on dust generation.

[0060] Furthermore, the system employs principal component analysis (PCA) to reduce the dimensionality of the three features. First, the original feature matrix is ​​standardized to have a mean of 0 and a variance of 1. Then, the covariance matrix is ​​calculated and eigenvalue decomposition is performed to extract principal components with a cumulative contribution rate greater than 90%. Typically, retaining 2-3 principal components is sufficient to cover most of the information from the original features. Through dimensionality reduction, the system effectively removes redundant information, reduces computational complexity, and simultaneously improves the model's sensitivity to key regulatory factors.

[0061] Specifically, in practical applications, this step is suitable for multi-equipment collaborative control scenarios at the tunnel face, especially under conditions of rapid changes in dust concentration and large fluctuations in work intensity. It can provide concise and representative input features for PID control, machine learning models, and LSTM prediction algorithms, thereby accelerating response speed and improving control accuracy. Its technical value lies in optimizing the data structure through feature engineering, providing high-quality data support for subsequent intelligent algorithms, and achieving efficient, energy-saving, and intelligent dust control.

[0062] S3 uses the feature dataset and calculated deviation data to generate basic adjustment quantities through the PID control algorithm. Combined with the optimal parameter combination output by the machine learning model, dynamic weight correction is performed to obtain the coordinated control parameters of the airflow direction adjustment quantity of the air distributor, the dust collection efficiency optimization quantity of the dust collection hood, and the airflow adjustment quantity of the negative pressure air duct.

[0063] Specifically, this step integrates the output results of PID control algorithms and machine learning models to generate coordinated control parameters for the airflow direction adjustment of the tunnel face air distributor, the dust collection efficiency optimization of the dust collection hood, and the airflow adjustment of the negative pressure ventilation duct. This is the core link in the entire intelligent ventilation and dust control system to achieve dynamic response and optimization decision-making.

[0064] Specifically, the system first calculates the deviation of real-time monitored parameters such as dust concentration, temperature, and wind speed based on a PID control algorithm, and generates basic adjustment quantities through three control links: proportional, integral, and derivative. The proportional coefficient Kp is typically set to 0.5~1.2, the integral time constant Ti is 120~300 seconds, and the derivative time constant Td is 10~30 seconds to ensure the system responds quickly while avoiding overshoot. The adjustment quantity output by the PID serves as a preliminary control command, guiding the adjustment of the distributor angle, the negative pressure of the dust collection hood, and the airflow of the negative pressure ventilation duct.

[0065] Furthermore, the system incorporates machine learning models (such as neural networks) for training and inference on historical and real-time data. Model inputs include key features such as dust concentration, wind speed, airflow pressure difference, and operational intensity coefficients, with the output being the optimal parameter combination for each device. In some implementations, the model employs standardization to map parameters of different dimensions to the [0,1] interval and extracts temporal features through a sliding window mechanism. The model inference results are weighted and fused with the PID output, with the weight coefficients dynamically adjusted according to the current operating conditions. For example, when dust concentration changes abruptly, the weight of the machine learning model can be increased to 0.7~0.9 to enhance predictive control capabilities.

[0066] Specifically, this step is widely applicable to the complex and ever-changing working environment of coal mine tunneling faces. Especially in high-dust operations such as coal cutting and blasting, the system can predict dust change trends in advance and generate coordinated control commands to achieve the linkage response of the air distributor, dust collection hood and negative pressure ventilation duct, thereby improving dust collection efficiency, maintaining airflow balance and reducing energy consumption.

[0067] Specifically, through the synergistic optimization of PID and machine learning, the system not only ensures real-time response capabilities but also possesses stronger adaptability and predictive capabilities, effectively solving the local imbalance problem caused by traditional independent control methods and improving the intelligence level and operating efficiency of the ventilation and dust control system.

[0068] Furthermore, S3 includes:

[0069] S31: The weight ratio of PID control quantity and machine learning model output parameter is dynamically adjusted according to the current work intensity coefficient. The higher the work intensity, the greater the weight ratio of the machine learning model.

[0070] Specifically, in some implementations, dynamically adjusting the weight ratio of the PID control input and the machine learning model output parameters based on the current workload intensity coefficient is one of the core strategies for achieving multi-algorithm collaborative control in this intelligent dust control system. This step introduces the workload intensity coefficient as a dynamic adjustment factor to achieve weighted fusion of the PID control and machine learning model output results, thereby optimizing control accuracy and system response speed under different operating conditions.

[0071] Specifically, the system first defines the Work Intensity Coefficient (WIC), which is calculated based on multi-source data such as the current work type at the tunneling face (e.g., coal cutting, blasting), equipment operating status (e.g., tunneling machine speed, advance speed), and dust generation rate. The WIC value range is typically set to [0,1], where 0 represents no work or low-intensity work, and 1 represents high-intensity continuous work. In the control logic, WIC is input as a regulation factor to the weighted fusion module, which linearly weights the PID control quantity (ΔPID) with the optimal parameters (ΔML) output by the machine learning model. This fusion strategy ensures that when the work intensity is low, the system primarily uses PID control to guarantee rapid response and stability; while during high-intensity work, the optimization capability of the machine learning model is enhanced to improve the overall efficiency of dust control.

[0072] Furthermore, the WIC update frequency is set to once every 30 seconds to ensure the real-time performance of the control strategy. The adjustment range of the PID control quantity is typically limited to ±10% of the equipment's rated output range, while the ΔML output of the machine learning model is normalized based on historical training data to make it comparable to the PID control quantity. In addition, the system sets a weight switching threshold for PID and ML; for example, when WIC ≥ 0.7, the ML weight ratio is not less than 60%, to adapt to operating conditions where dust concentration rises rapidly.

[0073] Specifically, this step is widely used in the dust-generating stages of operations such as coal cutting, loading, and transportation at the tunnel face. During these stages, the system optimizes the air curtain direction and airflow distribution in advance by increasing the weights of the machine learning model, thereby effectively suppressing dust diffusion and improving dust collection efficiency.

[0074] Specifically, this dynamic weighting mechanism significantly improves the system's adaptability and control accuracy. By combining the real-time response of PID control with global optimization through machine learning, the system can maintain dust concentration within a safe threshold even under complex operating conditions, while reducing energy consumption and improving the safety and intelligence level of tunneling operations.

[0075] S32 uses a fuzzy logic algorithm to perform nonlinear compensation on the corrected coordinated control parameters in order to eliminate control errors caused by equipment response delay.

[0076] Specifically, in some implementations, the nonlinear compensation of the corrected collaborative control parameters using fuzzy logic algorithms is a key step in achieving equipment response delay compensation and improving control accuracy in this intelligent dust control system. This step, based on the principle of Fuzzy Logic Control (FLC), dynamically corrects the collaborative control parameters output by the PID control algorithm and the machine learning model by constructing a fuzzy rule base and membership functions, thereby eliminating control errors caused by equipment response lag (such as adjustments to the air distributor angle or changes in the frequency of the negative pressure ventilation duct motor).

[0077] Furthermore, the fuzzy logic algorithm first receives adjustment parameters (such as airflow adjustment amount ΔQ, airflow direction adjustment angle Δθ) from PID control and machine learning models, and combines them with the current response state of the equipment (such as the frequency response delay time T_d of the fan motor, the angle change rate K_θ of the air distributor) as input variables. The input variables are fuzzified and mapped to corresponding linguistic variables (such as "low delay," "medium delay," and "high delay"), and quantized using triangular or trapezoidal membership functions. The fuzzy inference system performs multidimensional inference according to preset fuzzy rules (such as "if the delay is high and the deviation is large, the compensation coefficient should be increased"), and outputs a compensation coefficient α, which is used to weighted correct the original adjustment parameters.

[0078] Specifically, the input variables of fuzzy logic algorithms typically include deviation Δ (such as dust concentration deviation ΔC, temperature deviation ΔT), response delay time T_d (unit: seconds), and equipment adjustment rate K (such as angle change rate K_θ, unit: degrees / second). The compensation coefficient α is generally set between [0.8, 1.2] to ensure reasonable compensation and avoid over-adjustment that could cause system oscillations. The fuzzy rule base is constructed based on actual operating data, combining expert experience with system identification to form rules such as "IF ΔC is High AND T_d is Medium THEN α is 1.1", to achieve adaptive compensation for nonlinear dynamic responses.

[0079] Furthermore, this step is widely applied in tunnel face ventilation systems, particularly in operations such as adjusting the angle of the air distributor and changing the airflow of the negative pressure ventilation duct. Due to the complex and variable tunneling environment and the nonlinear lag characteristic of equipment response, traditional PID control struggles to achieve precise adjustment. By introducing a fuzzy logic compensation mechanism, the system can dynamically adjust control parameters under different work intensities and dust concentration trends, improving response speed and control accuracy, thereby effectively preventing excessive dust levels and localized ventilation imbalances.

[0080] Specifically, this step significantly improves the system's dynamic response capability and control robustness. Through nonlinear compensation, the system can more accurately predict and correct the deviation between the actual response of the equipment and the theoretical adjustment, reducing control overshoot and steady-state error, and improving the stability and safety of dust concentration control. At the same time, this method enhances the algorithm's adaptability to complex operating conditions without increasing hardware complexity, demonstrating the innovation and practicality of this invention in the field of intelligent ventilation control.

[0081] S4. Using the LSTM time series prediction model to analyze historical and real-time data, predict the trend of dust concentration and temperature change. When the predicted value exceeds the safety threshold, calculate the advance control amplitude based on the prediction deviation and generate a pre-control instruction based on the collaborative control parameters as the execution value. Simultaneously adjust the air volume of the negative pressure duct and the opening angle of the air distributor to achieve airflow balance.

[0082] Specifically, this step involves using an LSTM (Long Short-Term Memory) time-series prediction model to model and predict environmental parameters at the tunnel face, thereby enabling early identification of dust concentration and temperature change trends. When the predicted value exceeds the safety threshold, the control amplitude is calculated based on the prediction deviation, and a pre-control command is generated to simultaneously adjust the air volume of the negative pressure ventilation duct and the opening angle of the air distributor, so as to achieve proactive response in airflow balance and dust control.

[0083] Specifically, the LSTM model first receives preprocessed time-series data as input, including dust concentration (unit: ) collected over the past 60 minutes at 1-minute time steps. Temperature (unit: °C), wind speed (unit: m / s), negative pressure ventilation duct air volume (unit: ...) Key parameters include the opening angle of the air distributor (unit: °). Simultaneously, the system integrates face operation plan data (such as time nodes for coal cutting, blasting, etc.) to enhance the model's sensitivity to changes in operation intensity. The model employs a two-layer LSTM structure, containing 64 and 32 neurons respectively, with a Dropout layer (scale 0.2) introduced after each layer to prevent overfitting and improve generalization ability. The output layer is a fully connected layer, outputting the predicted dust concentration values ​​for each zone over the next 15-30 minutes, forming a continuous variation curve.

[0084] Furthermore, the system sets the dust concentration safety threshold to be... (According to the "Coal Mine Safety Regulations"), the temperature threshold is 30℃. If the dust concentration and temperature values ​​output by the prediction model exceed this threshold at any time point, the control mechanism will be triggered. The prediction deviation is calculated by taking into account the difference between the current value and the predicted value, and combining this with the time step (1 minute) and the degree of threshold deviation, using a linear scaling factor (e.g., 0.8-1.2) to calculate the advance control magnitude. For example, if the predicted dust concentration will reach [a certain value] in 5 minutes... The system can then initiate airflow adjustment 3 minutes in advance to control the actual concentration at [specific level]. the following.

[0085] Specifically, this step applies to dynamic ventilation and dust control scenarios in coal mine tunneling faces. Before high-dust operations such as coal cutting or blasting, the system uses an LSTM prediction model to identify the upward trend of dust concentration in advance and coordinates the adjustment of airflow and direction between the negative pressure ventilation duct and the air distributor. For example, when a rise in dust concentration is predicted, the system can adjust the airflow of the negative pressure ventilation duct from the current... Upgraded to The opening angle of the air distributor was adjusted from 30° to 45° to enhance local ventilation.

[0086] Specifically, this step, by introducing an LSTM prediction model, achieves a shift from passive response to active control, significantly improving the system's prediction accuracy and response speed to changes in dust concentration and temperature. Combined with PID control and a pressure balance algorithm, the system can optimize and adjust airflow and direction before prediction deviations occur, effectively preventing dust exceedances and ensuring the health and safety of workers. Simultaneously, by adjusting in advance rather than operating at full load, the system improves control performance while optimizing energy consumption, thus enhancing overall operating efficiency.

[0087] Furthermore, S4 includes:

[0088] S41, construct a neural network model containing two layers of LSTM units, where the first layer contains 64 neurons and the second layer contains 32 neurons, and set a Dropout layer with a ratio of 0.2 between the hidden layers to prevent overfitting.

[0089] Specifically, this step involves constructing a neural network model containing two layers of LSTM units to achieve time-series prediction of dust concentration at the tunnel face, thereby supporting dynamic control and early intervention of the system. In some implementations, this model employs a Long Short-Term Memory (LSTM) network structure to capture the complex dependencies of environmental parameters at the tunnel face over time, improving prediction accuracy and system responsiveness.

[0090] Furthermore, the first LSTM layer contains 64 neurons, responsible for extracting long-term dependency features from the input time-series data; the second LSTM layer contains 32 neurons, used to further abstract and compress the feature space, improving the model's generalization ability. A Dropout layer with a ratio of 0.2 is set between the two LSTM layers to randomly discard the outputs of some neurons, preventing the model from overfitting to the training data during training and enhancing its robustness under unknown conditions. A Dropout ratio of 0.2 means that in each training iteration, 20% of the neuron outputs are set to zero, thus forcing the model to learn more stable feature representations.

[0091] Specifically, the model receives time-series data from the past 60 minutes at 1-minute time steps, including key parameters such as dust concentration, temperature, wind speed, distributor angle, and negative pressure duct airflow. The data undergoes preprocessing, including linear interpolation to fill in short-term gaps, the mean of the first 30 minutes to replace long-term gaps, and 3-step moving average denoising to ensure the continuity and stability of the input data. The input layer organizes this data into a time-series matrix of shape (60, n), where n is the feature dimension.

[0092] Furthermore, the model output is the predicted dust concentration for each partition within the next 15 minutes. The output layer uses a fully connected structure with an output dimension of 15×m (where m is the number of partitions). During training, the model uses mean squared error (MSE) as the loss function, and the optimizer can be either Adam or RMSprop. The learning rate is typically set to 0.001, the batch size is 32, and the number of training epochs is controlled between 50 and 100 to achieve a balance between model convergence and overfitting.

[0093] Specifically, the LSTM model plays a crucial predictive role in the system, providing forward-looking data support for subsequent PID control and pressure balance algorithms. By predicting dust concentration trends in advance, the system can trigger low-risk pre-adjustment of airflow or high-risk collaborative control mechanisms to achieve proactive dust concentration control and improve the intelligence level and response efficiency of the tunnel face ventilation system.

[0094] S42 uses a 3-time-step moving average algorithm to denoise the preprocessed time-series data. At the same time, it uses linear interpolation to complete short-term missing data and uses the average of the first 30 minutes to replace long-term missing data.

[0095] Specifically, a 3-time-step moving average algorithm is used to denoise the preprocessed time-series data. Its technical principle is based on a sliding window mean filtering mechanism, aiming to smooth random noise in the data and improve the stability and reliability of environmental parameters (such as dust concentration, temperature, wind speed, etc.). The specific operation is as follows: taking the current time point as the center of the window, the data points from the two previous time steps are taken, and the data from the current time point, for a total of three time steps, are averaged to obtain the denoised output value for the current time point.

[0096] Specifically, at the parameter level, the sliding window length is set to 3 time steps (i.e., 3 minutes). This parameter has been proven in actual tests to effectively suppress high-frequency noise while preserving data trend characteristics. For example, under conditions of drastic dust concentration fluctuations, the 3-time-step moving average can reduce the noise amplitude by approximately 30%-40%, significantly improving the input quality of subsequent prediction models (such as LSTM). Furthermore, this algorithm does not rely on complex model training and only requires simple arithmetic operations, thus demonstrating good applicability in resource-constrained industrial control systems.

[0097] Furthermore, this denoising step is primarily used in the input data preprocessing stage of the LSTM prediction model. Due to the complex working environment at the tunnel face, sensor data is easily affected by factors such as dust interference and equipment vibration, leading to data anomalies or missing data within a short period. Through the moving average algorithm, the system can effectively eliminate such interference, providing high-quality data support for subsequent airflow regulation and linkage control.

[0098] Specifically, by reducing data noise, the accuracy of environmental parameter predictions is improved, thereby enhancing the system's ability to predict changes in dust concentration. By combining linear interpolation with the mean value of the previous 30 minutes as a replacement strategy, the system can maintain the continuity and stability of the prediction model even when faced with missing data, ensuring the timeliness and effectiveness of control commands and laying a solid foundation for the intelligent operation of the ventilation and dust control system at the tunnel face.

[0099] S5, when the system detects that the global pressure balance coefficient is in a severely unbalanced state, executes a three-step coordinated control process:

[0100] Specifically, when the system detects a severe imbalance in the global pressure balance coefficient, a three-step coordinated control process will be initiated to achieve rapid recovery and stable operation of the tunnel face ventilation system. This process is based on the dynamic linkage mechanism between various ventilation devices (air distributors, dust hoods, and negative pressure ventilation ducts) in the system, combined with real-time monitoring data and predictive model output, to ensure reasonable airflow distribution, efficient dust emission, and stable temperature control.

[0101] Specifically, the system first collects real-time air pressure data from key nodes of the air distributor, dust collection hood, and negative pressure duct using a pressure differential sensor, and standardizes it to the [0,1] range to eliminate dimensional differences. The global pressure balance coefficient is calculated by weighted averaging of the relative air pressure states of each component, used to quantify the overall pressure balance of the system. When this coefficient is lower than a preset severe imbalance threshold (e.g., less than 0.3), the system determines it to be in a severe imbalance state, triggering a three-step collaborative control process: First, the system prioritizes increasing the airflow of the dust collection hood to enhance the local negative pressure effect and accelerate dust collection and emission. In this stage, the basic airflow adjustment is calculated using a PID control algorithm, and combined with the current collection efficiency of the dust collection hood, if it is lower than the minimum requirement (e.g., less than 80%), an efficiency compensation is further introduced to optimize the dust collection effect; Second, the system dynamically adjusts the airflow output of the negative pressure duct based on the current total airflow demand and the airflow change after the dust collection hood adjustment. The adjustment amount must meet the rated airflow range of the duct (e.g., The system first adjusts the airflow velocity within a safe range (e.g., 1.5–4.0 m / s) using feedback from a wind speed sensor. The second step involves fine-tuning the opening angle and blade direction of the air distributor to calibrate the remaining air pressure deviation. The adjustment angle range is typically 0°–90°, and a sliding average method (e.g., 3 time steps) is used to smoothly evaluate the adjustment effect, preventing frequent fluctuations from affecting system stability.

[0102] Furthermore, the system sets a severe imbalance threshold and an airflow adjustment step size (such as...). Key parameters such as wind speed safety range and dust hood efficiency lower limit are used to ensure the accuracy and safety of the control process. This step plays a crucial role in emergency response and system recovery in the entire intelligent control method, effectively preventing dust accumulation and ventilation failure caused by local imbalances, and improving the safety of the tunneling operation environment and the efficiency of system operation.

[0103] S51 prioritizes increasing the local negative pressure by increasing the airflow of the dust collection hood.

[0104] Specifically, in some implementations, when the system detects abnormal differences in dust concentration or airflow pressure in a localized area of ​​the tunnel face, it preferentially increases the airflow of the dust collection hood to raise the local negative pressure in that area, thereby enhancing dust collection efficiency and improving airflow organization. This step is technically implemented based on a dynamic adjustment mechanism for the dust collection hood's negative pressure system, combined with real-time monitoring data and predictive model outputs to achieve precise control.

[0105] Specifically, the system first collects the airflow pressure difference between the dust collection hood opening and the working surface in real time using a pressure differential sensor, and standardizes this pressure difference to the [0,1] range for easy algorithm processing. When the pressure difference is lower than a preset negative pressure threshold (e.g., 0.85), it indicates that the dust collection hood's collection capacity is insufficient, and the system will trigger an airflow adjustment mechanism. At this time, the controller sends a frequency adjustment command to the dust collection hood's fan motor based on the optimal parameters output by a preset PID control algorithm or machine learning model, gradually increasing the fan speed through frequency conversion control, thereby increasing the airflow. In actual operation, airflow adjustment is usually based on... The step size is adjusted progressively to ensure a smooth system response and avoid secondary dust generation caused by airflow disturbance.

[0106] Furthermore, the system sets the minimum effective negative pressure of the dust collection hood to -150 Pa, and the maximum airflow does not exceed its rated value (e.g., This prevents equipment overload. Simultaneously, by combining real-time data from dust concentration sensors, the system can set the dust concentration change rate (e.g., ≥). The system uses this as a trigger condition to further optimize the timing and magnitude of adjustments. Furthermore, to improve prediction accuracy, the system employs an LSTM model to predict dust concentration trends over a 15-minute period. If the predicted value approaches or exceeds a safety threshold (e.g., ...), the system will adjust the adjustment timing and magnitude accordingly. If so, the air volume of the dust collection hood should be increased in advance to achieve proactive prevention and control.

[0107] Specifically, this step is particularly suitable for scenarios where there is a sudden increase in dust concentration during tunneling operations, such as coal cutting and blasting, which are high-dust-generating stages. By prioritizing the increase in the airflow of the dust collection hood, the system can create a stronger negative pressure collection field before the dust spreads, effectively controlling the dust diffusion path and improving dust removal efficiency.

[0108] Specifically, by prioritizing the enhancement of the negative pressure capture capability of the dust collection hood, the system can achieve rapid response and efficient dust removal in local areas, while providing a stable basic airflow environment for the coordinated control of subsequent air distributors and negative pressure ducts, thereby improving the dynamic balance capability and dust control efficiency of the overall ventilation system.

[0109] S52 dynamically adjusts the frequency of the negative pressure fan motor according to the total air volume requirement.

[0110] Specifically, this step, "dynamically adjusting the frequency of the negative pressure ventilation duct motor based on the total air volume demand," is one of the core control links in the intelligent ventilation system at the tunnel face, based on the multi-source environmental parameter sensing and multi-algorithm fusion control. Its technical implementation principle lies in dynamically calculating the current and future total air volume requirements by real-time collection and analysis of key parameters such as dust concentration, temperature, wind speed, and pressure difference at the tunnel face, combined with PID control algorithms, machine learning models (such as LSTM), and pressure balance algorithms. Based on this, the operating frequency of the negative pressure ventilation duct motor is adjusted, thereby achieving precise air volume regulation and optimized control of system energy consumption.

[0111] Specifically, the system first synchronously collects dust concentration (unit: ) data from each zone of the tunnel face through a distributed sensor network. The data includes air velocity (m / s), temperature (°C), and inlet and outlet air pressures (Pa) of the air distributor, dust collection hood, and negative pressure duct, and calculates the pressure difference between each area. After standardization, this data is input into the PID controller and LSTM prediction model. The PID controller calculates the current dust concentration based on the set safety threshold (e.g., the upper limit of dust concentration specified in the "Coal Mine Safety Regulations" is 10). The deviation between the two is used to calculate the basic air volume adjustment; the LSTM model is based on the time series data of the past 60 minutes (the input window is 60 time steps, each step is 1 minute) to predict the changing trend of dust concentration in each zone in the next 15 minutes, and is corrected by combining the work intensity coefficient (such as coal cutting speed, blasting frequency) to generate the predicted air volume demand.

[0112] Furthermore, the system weights and fuses the PID control amount with the airflow demand output from the prediction model to calculate the total airflow control amount. This control amount must meet the frequency adjustment range of the negative pressure ventilation duct motor (typically 20Hz to 50Hz, corresponding to an airflow adjustment range of...). to The system takes into account the motor's rated power and the inverter's response time (typically 0.1 to 0.5 seconds). Finally, the system converts the adjusted target airflow value into a PWM signal executable by the inverter, which is then sent to the negative pressure ventilation duct motor via the PLC control system to achieve real-time dynamic frequency adjustment.

[0113] Specifically, it is suitable for the complex ventilation environment of coal mine tunneling faces, especially during high-dust operations such as coal cutting and blasting. The system can predict the rising trend of dust concentration in advance and proactively increase the air volume before it exceeds the standard, thereby effectively preventing dust accumulation and ensuring the breathing safety of workers. At the same time, through linkage control with air distributors and dust collection hoods, the system can synchronously adjust the airflow direction and collection efficiency, achieving synergistic optimization of global airflow balance and dust control.

[0114] Specifically, in terms of technical effectiveness, this step significantly improves the response speed and control accuracy of the ventilation system, and the dust concentration control error can be controlled within a certain range. Within this range, the airflow adjustment response time is less than 5 seconds, and the overall system energy consumption is reduced by approximately 20% to 30% compared to traditional full-load operation. Through dynamic frequency adjustment, not only is the dust control efficiency improved, but the equipment lifespan is also extended, demonstrating the innovative value of this invention in terms of intelligence, energy saving, and safety.

[0115] S53, fine-tune the opening angle of the air distributor to calibrate the remaining air pressure deviation, and repeat the above three-step process every 30-second feedback cycle until the system pressure is balanced.

[0116] Specifically, the step of "fine-tuning the opening angle of the air distributor to calibrate the remaining air pressure deviation, and repeating the above three-step process every 30-second feedback cycle until the system pressure is balanced" is a key control link in achieving dynamic pressure balance of the ventilation system in this intelligent dust control system. Its technical implementation is based on real-time feedback and closed-loop control principles, combined with PID control algorithms, pressure difference monitoring, and equipment linkage mechanisms to ensure the uniformity and stability of airflow distribution in various areas of the tunnel face.

[0117] Furthermore, the system uses a pressure differential sensor to collect the pressure values ​​of the air ducts before and after the air distributor in real time and calculates the pressure difference ΔP. This pressure difference serves as a quantitative indicator of whether the system is in a pressure balance state. An optimal pressure range is typically set (e.g., ΔP ∈ [0.1 kPa, 0.3 kPa]). Exceeding this range is considered a pressure deviation. The system compares ΔP with the set value and calculates the remaining pressure deviation ΔP_res, which serves as the basis for adjusting the air distributor opening angle.

[0118] Furthermore, the opening angle adjustment of the air distributor adopts a PID control strategy. The proportional term (P) directly outputs the adjustment amount based on the current deviation magnitude, the integral term (I) is used to eliminate long-term accumulated static deviations, and the derivative term (D) predicts the trend based on the deviation change rate to prevent overshoot. The PID output adjustment amount, after amplitude limiting, is converted into a control signal for the air distributor actuator (such as an electric actuator or hydraulic adjustment device), enabling continuous fine-tuning of the angle.

[0119] Optionally, during the adjustment process, the system will also combine the dust collection efficiency of the dust collection hood with the air volume output of the negative pressure duct to perform multi-device collaborative optimization. For example, when the dust collection efficiency of the dust collection hood decreases, the system will simultaneously adjust the angle of the air distributor to guide more airflow into the dust collection area and improve the dust collection effect.

[0120] Specifically, this step is repeated every 30 seconds in a feedback cycle, forming a closed-loop control system. Through high-frequency data acquisition and parameter adjustment, the system can quickly respond to airflow imbalances caused by changes in the intensity of work at the tunnel face and fluctuations in equipment operation, ensuring that key parameters such as dust concentration, temperature, and wind pressure remain within safe threshold ranges. This mechanism not only improves the dynamic response capability of the ventilation system but also effectively reduces energy consumption and enhances the system's intelligence and stability, providing a crucial guarantee for achieving efficient dust control and a safe working environment.

[0121] An embodiment of the present invention provides a linkage control method for ventilation equipment at a tunnel face based on pressure balance, which can realize the coordinated linkage and dynamic adjustment of ventilation equipment at the tunnel face, effectively improve the real-time performance and accuracy of dust control and temperature regulation, while optimizing system energy consumption and ensuring the safety of the working environment and ventilation efficiency.

[0122] Example 2

[0123] To achieve the above embodiments, this embodiment also provides a pressure balance-based linkage control system for tunnel face ventilation equipment, such as... Figure 2 As shown, it includes:

[0124] This intelligent dust control system incorporates various data acquisition devices (such as dust sensors, temperature sensors, and wind speed sensors), a data analysis system, and an automatic control system to monitor and analyze environmental parameters at the tunneling face in real time. It also enables the coordinated control of wind speed, air volume, and wind direction in the compressed air pipeline (including air distributors and dust collection hoods) and negative pressure ventilation ducts, thereby efficiently controlling the dust concentration and temperature at the tunneling face.

[0125] S101, Multi-sensor Data Acquisition and Intelligent Algorithm Innovation: The system utilizes dust sensors, temperature sensors, and wind speed sensors to collect real-time environmental data from the working surface, which is then analyzed and processed by a data analysis system. Intelligent algorithms (including PID control, machine learning algorithms, and predictive models) automatically adjust parameters such as wind speed, airflow, and air direction based on real-time data, thereby ensuring that the dust concentration at the working surface is controlled within a safe range and that the temperature remains stable. Specifically, it includes the following sub-steps:

[0126] S1011, PID control algorithm data processing procedure: Data input and deviation calculation: First, real-time environmental parameters of the tunneling face are collected, including dust concentration, temperature, and wind speed; at the same time, safety thresholds for each parameter are preset. The real-time collected parameter values ​​are subtracted from the corresponding safety thresholds to obtain the dust concentration deviation, temperature deviation, and wind speed deviation, thereby clarifying the gap between the current environmental parameters and the safety standards.

[0127] Furthermore, the PID control input is calculated through three stages: proportional (P), integral (I), and derivative (D), based on the calculated deviation value. The proportional stage directly outputs the adjustment amount according to the deviation magnitude, amplifying the current deviation's control effect on the equipment. The integral stage accumulates historical deviations, eliminating long-term static deviations and preventing parameters from continuously deviating from the threshold. The derivative stage analyzes the deviation's changing trend, predicts its future direction, and adjusts the control intensity in advance to prevent overshoot. The results of the three stages are then superimposed to obtain the final PID control input. The output control command converts the calculated PID control input into parameters executable by the equipment.

[0128] S1012, Machine Learning Algorithm Data Processing: Dataset Construction and Preprocessing: On the one hand, historical operating data is collected, including records of dust concentration, distributor angle, negative pressure duct airflow, system energy consumption, etc., under different operating conditions; on the other hand, the current real-time collected and preprocessed environmental parameters are obtained, and standardized methods are used to separate parameters of different dimensions (such as dust concentration). The system maps temperatures (in °C) to a unified range to eliminate the influence of dimensions. It also identifies key features that significantly impact control effectiveness, such as dust change rate, airflow pressure difference, and work intensity coefficient, to construct a complete dataset. Model training and inference: A neural network machine learning model is used, taking key features from the historical dataset as input and dust control effectiveness (e.g., whether dust concentration meets standards) as output. The model is trained to learn the mapping relationship between features and control effectiveness. After training, the pre-processed key features are input into the model, which then calculates and outputs the optimal parameter combination for the distributor, dust collection hood, and negative pressure duct under the current operating conditions. Parameter optimization output: The optimal parameter combination output by the machine learning model is compared with the initial equipment adjustment parameters output by the PID control algorithm. Weighting coefficients are set according to the actual operating conditions to correct the initial PID adjustment parameters. For example, if the machine learning model determines that a certain area requires a larger airflow, the corresponding PID airflow adjustment amount for that area can be appropriately increased. Ultimately, a coordinated control parameter that balances real-time performance and optimal effect is obtained, and equipment control commands are generated.

[0129] S1013, Predictive Model Algorithm (LSTM) Data Processing: Time Series Data Preprocessing: Time series data recorded at fixed time steps within the past hour is collected, including dust concentration, temperature, wind speed, and equipment operating parameters (distributor angle, negative pressure ventilation duct air volume). Simultaneously, tunneling face operation plan data (such as blasting and coal cutting operation time nodes) is entered. LSTM Model Prediction: The preprocessed time series dataset is input into the trained LSTM model. The model captures the long-term and short-term dependencies in the time series data through its internal memory units, such as the pattern of dust concentration changes with operation intensity and equipment parameters. Based on this dependency, the model predicts the dust concentration and temperature change trends for a future period (e.g., 15-30 minutes), outputting the predicted dust concentration and temperature values ​​at each time point within the future period, forming a complete change curve. Application of Prediction Results: The predicted future parameter values ​​output by the model are compared with preset safety thresholds. If the predicted dust concentration or temperature will exceed the safety threshold at a certain future moment, an early control mechanism is triggered. Based on the difference between the predicted value and the threshold, the magnitude of advance control is calculated, the air volume of the negative pressure duct is increased in advance, the airflow direction of the air distributor is adjusted, and a pre-control instruction is generated and sent to the equipment.

[0130] S102, Interlocking Control of Air Distributor and Dust Collection Hood: The system optimizes airflow distribution through the interlocking control of the air distributor and dust collection hood. When the dust concentration in a certain area exceeds the standard, the algorithm will control the air distributor to adjust the airflow direction, guiding more airflow into that area, increasing the air volume in that area, and carrying away excess dust. At the same time, the airflow direction of the dust collection hood will be adjusted to optimize the dust collection efficiency. When the temperature exceeds the standard, the system will increase the air supply volume to promote heat dissipation and prevent high temperatures from affecting the working environment.

[0131] S103, Negative Pressure Duct Airflow and System Linkage Optimization: The system innovatively links the airflow adjustment of the negative pressure duct with other ventilation devices such as air distributors and dust collection hoods. By monitoring the pressure difference of the airflow at the working face in real time and combining sensor data feedback, the system can dynamically adjust the airflow of the negative pressure duct to ensure airflow balance and improve dust control efficiency. Specifically, this includes the following data processing procedures:

[0132] The negative pressure ventilation duct airflow regulation linkage algorithm first synchronously collects the dust concentration in each zone of the tunnel face, the inlet and outlet air pressure of the distributor (calculating the pressure difference), the current airflow of the negative pressure ventilation duct, and the dust collection efficiency of the dust collection hood. The dust concentration and pressure difference are standardized to the [0,1] range. Then, the global deviation is calculated using the dust concentration safety threshold. The basic airflow regulation amount is obtained through a PID algorithm. If the pressure difference deviates from the optimal range, a pressure correction amount is calculated. If the dust collection efficiency of the dust collection hood is lower than the minimum requirement, an efficiency compensation amount is calculated. The three are combined into a total regulation amount, and the final airflow is obtained according to the upper and lower limits of the ventilation duct's rated airflow. This is converted into a motor frequency signal and synchronously sent to the distributor control system to adjust the opening angle, achieving multi-device linkage, ensuring airflow balance and dust control effectiveness, while avoiding system imbalance and airflow waste. The algorithm includes the following sub-steps:

[0133] S1031: The system collects parameters such as dust concentration, air pressure difference, and air volume at the tunneling face in real time and performs normalization processing. Using a dust safety threshold as the target, a PID control algorithm calculates the basic air volume adjustment. When the pressure difference deviates from the optimal range or the dust collection hood efficiency falls below the minimum value, air volume correction and compensation are performed respectively. After comprehensive adjustment, the air volume is converted into a motor frequency signal, synchronously controlling the air distributor to adjust the opening angle to maintain airflow balance and dust concentration within a safe range. Specifically, S1031 may include:

[0134] S10311, synchronously collect the following parameters: dust concentration in each zone of the tunnel face. air pressure at the outlet of the air distributor With inlet air pressure Calculate the pressure difference Current air volume of negative pressure ventilation duct Dust collection hood collection efficiency feedback value (Calculated from the dust concentration difference between 0 and 1), the parameters are mapped to the [0,1] interval to eliminate the influence of dimensions: , .

[0135] S10312, based on the dust concentration safety threshold Calculate the global dust deviation as the target: ,pass The algorithm outputs the basic airflow adjustment amount: .

[0136] S10313, when pressure difference Deviation from optimal range hour( Calculate the correction amount (to account for allowable fluctuations): .

[0137] S10314, if the dust collection efficiency of the dust collection hood Increase airflow compensation: .

[0138] S10315, combining PID basic adjustment, pressure correction, and efficiency compensation, yields the total airflow regulation: .

[0139] S10316 converts the final air volume into a frequency adjustment signal for the air duct motor and sends it synchronously to the air distributor control system, so that it can adjust the opening angle according to the air volume change to ensure airflow balance.

[0140] S1032, a pressure differential sensor monitors the airflow pressure difference before and after the air distributor in real time. If the detected pressure difference is too large (e.g., below a preset value), the system will increase the overall airflow by increasing the air volume of the negative pressure duct, thereby accelerating the emission of high-concentration dust. Wind speed and air volume linkage control: The system combines the air volume of the negative pressure duct with data such as wind speed and air volume from other parts of the working surface, and uses a multi-objective optimization algorithm to ensure airflow coordination between the negative pressure duct, air distributor, dust collection hood, and other equipment. For example, when the wind speed is too low and the dust concentration is high, the system will increase the air volume of the negative pressure duct and guide the airflow to the high-dust area by controlling the opening and blade angle of the air distributor. Predictive and dynamic adjustment mechanism: By introducing a machine learning prediction model (LSTM), the system can predict the airflow requirements of the negative pressure duct, air distributor, and dust collection hood. The model uses the fusion of historical and real-time data to predict future dust concentration trends. Based on the prediction results, the system adjusts the airflow of the negative pressure ventilation duct in advance and optimizes the airflow distribution of other equipment to avoid sudden dust concentration exceeding the standard.

[0141] Specifically, the predictive model (LSTM) first collects data on the tunnel face environment, equipment operation, and work status at 1-minute intervals. Short missing data is filled with linear interpolation, long missing data is replaced with the average of the previous 30 minutes, and noise is reduced using a 3-time-step moving average. Then, a two-layer LSTM network is trained using the past 60 minutes of data as the input window. Subsequently, the current 60 minutes of preprocessed data is input into the model to obtain predicted dust concentration values ​​for each zone in the next 15 minutes. The prediction results are then corrected based on the work intensity, and compared with safety thresholds to trigger tiered measures such as low-risk pre-adjustment of airflow and high-risk coordinated control, achieving early dust concentration prevention and control. The specific process is as follows:

[0142] S10321, Data Input: The input data contains three types of features, forming a multidimensional time series: Environmental parameters: dust concentration in each zone. Negative pressure ventilation duct air volume Dust collection hood angle Operating status parameters: Tunneling machine operating power Number of workers .

[0143] S10322, Data cleaning: Short-term missing data is filled with linear interpolation, and long-term missing data is replaced with the average of the previous 30 minutes. Noise is filtered by moving average with 3 time steps. The time series is constructed with a step size of 1 minute, using the past 60 minutes of data as the input window, to predict the dust concentration of each zone in the next 15 minutes.

[0144] Specifically, the model structure is as follows: Input layer: receives a 60-minute time series matrix; Hidden layers: two LSTM units (containing 64 and 32 neurons respectively), with a 0.2 dropout ratio to prevent overfitting; Output layer: a fully connected layer outputting a 15-minute dust concentration prediction matrix for each region; Training parameters: the training and validation sets are divided in an 8:2 ratio, using time series cross-validation, with the root mean square error (RMSE) as the loss function, as shown in the formula: ,in For predicted values, This is the actual value. Number of zones; Real-time prediction: Input the current 60 minutes of preprocessed data into the trained model, and output the predicted dust concentration values ​​for each zone for the next 15 minutes. And calculate the 95% confidence interval. ,in This represents the standard deviation of the historical prediction error for this partition.

[0145] Specifically, for example, in areas with high dust concentration, the system can predict through a predictive model that the dust concentration in the area will further increase in the future. Therefore, it can increase the air volume of the negative pressure duct in advance, and increase the air supply of the air distributor and dust collection hood to alleviate the dust pollution problem in advance.

[0146] S1033, the linkage control and pressure balance algorithm, includes the following sub-steps:

[0147] S10331, Data Acquisition and Preprocessing: Collect parameters of the air distributor (air pressure and opening angle of each air outlet), dust collection hood (negative pressure at the hood opening and air volume captured), and negative pressure air duct (air pressure at the inlet and outlet and current air volume) at 1-minute intervals, and standardize the air pressure of each component to a relative air pressure in the 0-1 range to eliminate the influence of dimensions.

[0148] S10332, to quantify the pressure balance of the entire ventilation system, the algorithm introduces a global pressure balance coefficient. By comprehensively considering the relative wind pressure status of all key components, it determines whether the system is unbalanced. The calculation formula is as follows: Where N is the total number of components involved in the pressure calculation. For balance, Mild imbalance, For severe imbalance; for mild imbalance adjustment: only the air distributor and negative pressure duct are linked. First, the angle adjustment of the air distributor is calculated to correct the abnormal wind pressure. Then, the air volume of the duct is adjusted synchronously to compensate for the change in wind resistance and avoid fluctuations in other components. For severe imbalance adjustment: three-step coordinated control is used: first, the air volume of the dust collection hood is increased to improve the negative pressure. Then, the air volume of the duct is supplemented to adapt to the total demand. Finally, the angle of the air distributor is finely adjusted to calibrate the remaining wind pressure deviation.

[0149] S10333 converts the adjustment amount into a device command, and recalculates the parameters 30 seconds after sending it. If the parameters meet the standard, they are maintained; if not, adjustments are repeated until the system pressure is balanced. The airflow adjustment of the negative pressure duct will work in coordination with other ventilation components on the working face, such as air distributors and dust collection hoods, through a pressure balance algorithm. The system monitors the pressure changes of the entire system in real time. If the pressure in a certain part of the air duct is low or the pressure difference is too large, the system will automatically adjust the airflow of the negative pressure duct to ensure the overall airflow balance and avoid insufficient ventilation in local areas.

[0150] S104, Overall Effects and Applications: As a core component of intelligent dust control systems, negative pressure ventilation ducts play a crucial role in efficient dust control, stable ventilation, and safety assurance. Through coordinated control with devices such as air distributors and dust collection hoods, negative pressure ventilation ducts ensure airflow balance, effective dust emission, and temperature regulation, significantly improving the overall system efficiency. Furthermore, the system incorporates advanced algorithmic PID control, machine learning prediction models, and pressure difference algorithms to precisely adjust the airflow of the negative pressure ventilation ducts. This not only allows for rapid response to environmental changes but also optimizes energy consumption and enhances the system's intelligence level.

[0151] The pressure balance-based ventilation equipment linkage control system of the tunneling face according to this invention not only ensures the safety, comfort and efficiency of the working face environment, but also effectively reduces energy consumption, providing an intelligent and reliable ventilation and dust control solution for coal mine operations.

[0152] Example 3

[0153] To achieve the above embodiments, this embodiment also provides a pressure balance-based interlocking control device 10 for ventilation equipment in tunnel faces, such as... Figure 3 As shown, it includes:

[0154] The multi-source parameter acquisition module 100 is used to synchronously acquire multi-source environmental parameters and equipment operating parameters related to the tunnel face.

[0155] Specifically, dust sensors, temperature sensors, wind speed sensors, and differential pressure sensors are used for distributed data collection. The collected parameters are uploaded to the data analysis system in real time via a wireless transmission module, and edge computing nodes are used for preliminary data fusion processing.

[0156] The parameter standardization and feature construction module 200 is used to standardize the collected parameters to a uniform range, calculate the dust concentration deviation, temperature deviation and pressure difference deviation based on the preset safety threshold, and construct a feature dataset including dust change rate, airflow pressure difference and work intensity coefficient.

[0157] Specifically, parameters such as dust concentration, temperature, and wind speed are mapped to the [0,1] interval using a linear normalization method to eliminate dimensional differences; dust change rate, airflow pressure difference, and work intensity coefficient are selected as key features, and principal component analysis is used for feature dimensionality reduction.

[0158] The PID and machine learning collaborative control module 300 is used to generate basic adjustment quantities through the PID control algorithm using feature datasets and calculated deviation data. It combines the optimal parameter combination output by the machine learning model to perform dynamic weight correction, thereby obtaining collaborative control parameters for the airflow direction adjustment of the distributor, the dust collection efficiency optimization of the dust hood, and the airflow adjustment of the negative pressure duct.

[0159] Specifically, the weight ratio of PID control input and machine learning model output parameters is dynamically adjusted according to the current workload coefficient. The higher the workload, the greater the weight ratio of the machine learning model. The corrected collaborative control parameters are then compensated nonlinearly using a fuzzy logic algorithm to eliminate control errors caused by equipment response delay.

[0160] The LSTM time series prediction and pre-control module 400 is used to analyze historical and real-time data using the LSTM time series prediction model to predict the dust concentration and temperature change trends in the next 15-30 minutes. When the predicted value exceeds the safety threshold, the module calculates the pre-control magnitude based on the prediction deviation and generates a pre-control command based on the collaborative control parameters. The module simultaneously adjusts the air volume of the negative pressure duct and the opening angle of the air distributor to achieve airflow balance.

[0161] Specifically, a neural network model containing two layers of LSTM units is constructed, with the first layer containing 64 neurons and the second layer containing 32 neurons. Dropout layers with a ratio of 0.2 are set between the hidden layers to prevent overfitting. A 3-time-step moving average algorithm is used to denoise the preprocessed time series data. At the same time, linear interpolation is used to complete short-term missing data, and mean replacement is used for long-term missing data.

[0162] This invention discloses a pressure balance-based ventilation equipment linkage control device for tunneling faces. It synchronously collects multi-source environmental parameters such as dust concentration, temperature, and wind speed at the tunneling face using distributed sensors, and utilizes wireless transmission and edge computing to achieve real-time data fusion and preliminary processing. A parameter standardization and feature construction module normalizes the data and extracts key features, eliminating dimensional differences. A PID and machine learning collaborative control module dynamically adjusts control weights, combines fuzzy logic to compensate for equipment delays, and optimizes the control parameters of the air distributor, dust collection hood, and negative pressure ventilation duct. An LSTM time-series prediction model predicts future dust and temperature trends based on historical and real-time data, generating pre-control commands to achieve airflow balance and active control. This system significantly improves the accuracy, real-time performance, and robustness of environmental control, effectively ensuring safe production at the tunneling face.

[0163] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0164] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for coordinated control of ventilation equipment in a tunnel face based on pressure balance, characterized in that, include: S1, synchronously collects multi-source environmental parameters and equipment operating parameters related to the tunnel face; S2, after standardizing the collected parameters to a uniform range, calculate the dust concentration deviation, temperature deviation and pressure difference deviation based on the preset safety threshold, and construct a feature dataset including dust change rate, airflow pressure difference and work intensity coefficient. S3, using the feature dataset and calculated deviation data, generates a basic adjustment quantity through a PID control algorithm, and performs dynamic weight correction by combining the optimal parameter combination output by the machine learning model, to obtain coordinated control parameters for the airflow direction adjustment of the distributor, the dust collection efficiency optimization of the dust hood, and the airflow adjustment of the negative pressure duct; S3 also includes: S31, dynamically adjust the weight ratio of PID control quantity and machine learning model output parameters according to the current work intensity coefficient. The higher the work intensity, the greater the weight ratio of the machine learning model. S32, the corrected coordinated control parameters are nonlinearly compensated using a fuzzy logic algorithm to eliminate control errors caused by equipment response delay; S4, using an LSTM time-series prediction model to analyze historical and real-time data, predicts the trends of dust concentration and temperature changes. When the predicted value exceeds the safety threshold, the pre-control amplitude is calculated based on the prediction deviation, and a pre-control command based on the collaborative control parameters is generated to simultaneously adjust the negative pressure duct airflow and the air distributor opening angle to achieve airflow balance; S4 also includes: S41, construct a neural network model containing two layers of LSTM units, where the first layer contains 64 neurons and the second layer contains 32 neurons, and set a Dropout layer with a ratio of 0.2 between the hidden layers to prevent overfitting; S42 uses a 3-time-step moving average algorithm to denoise the preprocessed time-series data, and at the same time uses linear interpolation to fill in the missing data in the short time period.

2. The method as described in claim 1, characterized in that, S1 further includes: S11 uses a distributed array of dust sensors, temperature sensors, wind speed sensors, and pressure difference sensors to collect relevant multi-source environmental parameters and equipment operating parameters. The multi-source environmental parameters include: dust concentration, temperature, wind speed, pressure difference, air distribution fan opening angle, dust collection hood efficiency, and negative pressure duct air volume. S12 uploads the collected parameters to the data analysis system in real time via a wireless transmission module, and performs preliminary data fusion processing using edge computing nodes.

3. The method as described in claim 1, characterized in that, S2 further includes: S21, dust concentration, temperature and wind speed are mapped to the [0,1] interval through linear normalization to eliminate dimensional differences; S22, dust change rate, airflow pressure difference, and work intensity coefficient were selected as key features, and principal component analysis was used to perform feature dimensionality reduction.

4. The method as described in claim 1, characterized in that, Also includes: S5, when the system detects that the global pressure balance coefficient is in a severely unbalanced state, executes a three-step coordinated control process: S51 prioritizes increasing the local negative pressure by increasing the airflow of the dust collection hood; S52, dynamically adjusts the frequency of the negative pressure fan motor according to the total air volume requirement; S53, fine-tune the opening angle of the air distributor to calibrate the remaining air pressure deviation, and repeat the above three-step process within the feedback cycle until the system pressure is balanced.

5. A linkage control device for ventilation equipment in a tunneling face based on pressure balance, characterized in that, include: The multi-source parameter acquisition module is used to simultaneously acquire multi-source environmental parameters and equipment operating parameters related to the tunnel face; The parameter standardization and feature construction module is used to standardize the collected parameters to a uniform range, calculate the dust concentration deviation, temperature deviation and pressure difference deviation based on the preset safety threshold, and construct a feature dataset including dust change rate, airflow pressure difference and work intensity coefficient. The PID and machine learning collaborative control module is used to generate basic adjustment quantities through the PID control algorithm using feature datasets and calculated deviation data. It combines the optimal parameter combination output by the machine learning model to perform dynamic weight correction, thereby obtaining collaborative control parameters for the airflow direction adjustment of the air distributor, the dust collection efficiency optimization of the dust collection hood, and the airflow adjustment of the negative pressure duct. The PID and machine learning collaborative control module is also used to dynamically adjust the weight ratio of the PID control quantity and the machine learning model output parameters according to the current work intensity coefficient. The higher the work intensity, the greater the weight ratio of the machine learning model. The corrected coordinated control parameters are nonlinearly compensated using a fuzzy logic algorithm to eliminate control errors caused by equipment response delay; The LSTM time series prediction and pre-control module is used to analyze historical and real-time data using the LSTM time series prediction model to predict the trend of dust concentration and temperature changes. When the predicted value exceeds the safety threshold, the module calculates the pre-control amplitude based on the prediction deviation and generates a pre-control instruction based on the collaborative control parameters as the execution value. It simultaneously adjusts the air volume of the negative pressure duct and the opening angle of the air distributor to achieve airflow balance. The LSTM time series prediction and pre-regulation module is also used to construct a neural network model containing two layers of LSTM units, where the first layer contains 64 neurons and the second layer contains 32 neurons, and a Dropout layer with a ratio of 0.2 is set between the hidden layers to prevent overfitting; a 3-time-step moving average algorithm is used to denoise the preprocessed time series data, and linear interpolation is used to complete short-term missing data, while mean replacement is used for long-term missing data.

6. The apparatus as claimed in claim 5, characterized in that, The multi-source parameter acquisition module is also used for: Distributed point collection is carried out using dust sensors, temperature sensors, wind speed sensors and pressure difference sensors to collect relevant multi-source environmental parameters and equipment operating parameters. The multi-source environmental parameters include: dust concentration, temperature, wind speed, pressure difference and the opening angle of the air distributor, dust collection efficiency of the dust collection hood, and air volume of the negative pressure air duct. The collected parameters are uploaded to the data analysis system in real time via a wireless transmission module, and preliminary data fusion processing is performed using edge computing nodes.

7. The apparatus as claimed in claim 5, characterized in that, The parameter standardization and feature construction module is also used for: Dust concentration, temperature, and wind speed parameters are mapped to the [0,1] interval using a linear normalization method to eliminate dimensional differences; Dust change rate, airflow pressure difference, and work intensity coefficient were selected as key features, and principal component analysis was used to perform feature dimensionality reduction.

8. The apparatus as claimed in claim 5, characterized in that, Also includes: The pressure imbalance coordination and control module is used to execute a three-step coordination and control process when the system detects that the global pressure balance coefficient is in a severely imbalanced state: Firstly, increase the local negative pressure by increasing the airflow of the dust collection hood; The frequency of the negative pressure fan motor is dynamically adjusted according to the total air volume requirement. Fine-tune the opening angle of the air distributor to calibrate the remaining air pressure deviation, and repeat the above three-step process within the feedback cycle until the system pressure is balanced.

Images